ap539813/emotion-recognition

Detect the emotion from facial expressions with an accuracy of 81%.

EMOTION RECOGNITION (AI)

1.1 AIM OF THE PROJECT

To recognise the emotions of the human face using CNN algorithm.

• Create an own face dataset or to import an existing dataset from the internet.
• Create a model and train it to increase the accuracy for recognize the facial emotions.
• To use the trained model for emotions recognise in real time

1.2 WORKDONE

• Used 3 datasets - FER2013, JAFFE, Own face dataset used to train the model.
• The model has been developed with a maximum accuracy of 80.61%, with the help of python libraries
• The model is then used to recognise the emotions of a human face from the real time data using web camera

1.3 DATASETS

1.3.1 Fer2013 : It is a face-set uploaded by kaggle.com that contains 35887 datasets 1.3.2 JAFFE database : This was uploaded by kasrl.org that has 214 datasets 1.3.3 Our own database : We created our own database which has 360 datasets created using Viola-Jones method.

• Our own dataset
• jaffe dataset
• FER2013 dataset

1.4 Emotions recognized

1. Happy
2. Sad
3. Angry
4. Neutral
5. Scared
6. Surprise

1.5 DIFFICULTIES FACED

• The downloaded databases was extremely biased.
• Not easily accessible datasets.
• Creating our own data was quite difficult.

1.6 RESULTS:

• Happy face
• Sad face
• Angry face
• Neutral face
• Surprised face

watch the complete output here

https://youtu.be/jnS3BogBTtM

1.7 REFERENCE

Octavio Arriaga Et Al., “Real-time Convolutional Neural Networks for Emotion and Gender Classification” ,2017.